[Train] with NumWFOCycles = 16 produces 16 .c files in Zorro's Data/ dir.
I have cycle-specific models, and I would like Zorro to use them for [Test]. I figured I could put a cycle-specific predict() function in each of the .c data files, and the main run() function would call the right one depending on the cycle number. I am doing this because I want a more complex nnet than PERCEPTRON (more than one layer, with output layer larger than 1.)
In barium_strat.c I have
void predict(var *xrow, var *pcx, var *L0, var *L1, var *L2);
var xrow[95];
var pcx[95];
var L0[96];
var L1[253];
var L2[5];
# snip
function run() {
# snip
# Populate xrow.
predict(xrow, pcx, L0, L1, L2);
# snip
# Maybe enter some orders based on values in L2.
}
And in each Data/barium_strat_??.c file I have a
function predict(var *xrow, var *pcx, var *L0, var *L1, var *L2) {
# Populate pcx L0, L1, L2
}
For [Test] I see the following error during compile. Then the test runs, but the predict function is never called, and the values of the output variables pcx, L0, L1, L2 are always unmodified.
barium_strat compiling................................................
Undefined function: predict. assets....
Walk-Forward Test: barium_strat portfolio 2009..2013
Clearly I have some misunderstanding. If it's not too much trouble, can you point me in a better direction? Or am I trying to do something impossible?
